Tenchu San Portable major issues!

[wolfokami321]

Tenchu san portable is currently unplayable because ingame some of the levels have no collision loaded meaning you and the enemies can walk through walls... not only that but the in game loading screens are infinite.

Note: I've tested two different rips of the game along with my original.
To reiterate:
-missing unrendered collision?
-infinite ingame loading screen

[daniel229]

I don't see anyone walk through walls play on PC.The loading screen is an atrac3p issue.

[wolfokami321]

Im using the android version and there is no attrac option on ppsspp 9.9.1

[daniel229]

Any savestate to reproduce the missing unrendered collision problem?Atrac option won't help it,it need some change to the code.

[wolfokami321]

I can upload one but it doesn't matter because everytime I start the game over the same thing happens. Try the latest build for android ppsspp with the iso

[wolfokami321]

https://www.dropbox.com/s/egl78uxpgcs9o25/ULJM05505_1.01_0.ppst?dl=0

[wolfokami321]

Here it is

[daniel229]

It does not happen on PC,so it's an ARM specsific.I don't have android to test.Does disable JIT help the problem?

[wolfokami321]

No it doesn't help at all... can any other person try it on their android?

[wolfokami321]

Wait there's an android emulator you can download try that

[unknownbrackets]

What version of ppsspp are you using? Is it the latest git build? There have been some recent armjit fixes.

-[Unknown]

[wolfokami321]

Ok the build number is 9.9.1-1310 g1a0f537

[wolfokami321]

I tried the latest last night 9.9.1. 1318 and it still didn't fix anything

[wolfokami321]

Pc works fine on my end too except I get eternal in level loading aswell. I cannot disable attrac on pc either since there is no option

[wolfokami321]

No replies???

[wolfokami321]

Please fix these two issues, also kurenai portable has infinite loading aswell

[wolfokami321]

Also I have used the latest buildbot put out this morning to no avail

[hrydgard]

Does it work now, or why did you close it?

[wolfokami321]

Closed it because it wasn't getting any attention. .. ive been trying every build. What is the difference between the windows and android builds? They are both compiled from the same source right?

[hrydgard]

Not the JIT, which has to be done separately for ARM and x86. So it's very likely an ARM JIT bug.

By the way, you have way too little patience. People don't stop what they're doing just to work on the newest bug all the time.

[wolfokami321]

Bro I see at least 20 new builds a day XD im pretty sure you guys make a living doing this... Anyways is there any progress on making the Jit code for ARM(Android) the same as x86(windows)

[wolfokami321]

Also I disabled JIT and nothing changed, it probably means it isn't JIT related. Also, why is there no option to disable attrac because that could be what's causing in level loading issues?

[Bigpet]

@wolfokami321 so the collisions work on PC but they don't work on mobile, even if you disable JIT?

[wolfokami321]

Yes

[wolfokami321]

Now on pc and mobile the infinite loading is still there in level

[Bigpet]

Just assume that we still think it's broken unless we either ask you to retest or you tell us something changed. There's no need to remind us that it's still broken with this frequency.

[wolfokami321]

Also why was the attrac option removed, some games have terrible attrac related issues still

[wolfokami321]

Bump...

[hrydgard]

@sum2012 interesting, thanks! That might explain it.. I suppose we shouldn't enable that, or only when the PSP game requests flush-to-zero fp rounding...

[unknownbrackets]

Is the graphic issue Android only?

There's a setting in ppsspp.ini for flush to zero - it defaults to true on Android:
ForceFlushToZero = True

Does changing it to False improve anything? This will only impact jit, so if the problem is outside jit it won't make a difference (for example, interp will still use flush to zero.)

-[Unknown]

[hrydgard]

I think we probably should kill that setting, it's dangerous and I don't think there's enough speed difference to motivate it. Better to strive to be as accurate as reasonably possible with rounding..

[xiushudongfang]

@unknownbrackets ， unfortunately,I have tested it .It did not work.

[hrydgard]

Here's the documentation for the FPSCR register on Cortex A8 (and later, probably):

http://infocenter.arm.com/help/index.jsp?topic=/com.arm.doc.ddi0344b/Chdfafia.html

Mostly, it just masks away "DNM" bits (do not modify), the only relevant thing that is different after that commit is that it forces the rounding mode to NEAREST. So that might not be good then..

Another interesting thing is that both before and after, it doesn't just set the flush-to-zero bit but also the "DN" bit (Default NaN)... I wonder if that's good.

Anyway, @sum2012 and @xiushudongfang can you try this:

In FPU_SetFastMode, change 0x04086060 to 0x04c86060. This should keep the original rounding mode.

If that doesn't do it, I'll just revert it or remove SetFastMode entirely.

[unknownbrackets]

@hrydgard well, I kept the setting mostly to maintain old behavior on arm (since I didn't know how much it might affect other devices) and so we had another datapoint. I'm totally okay with killing the setting.

Do we have any numbers on actual improvements SetFastMode even does?

I should note that some tweaks to the armjit SetRoundingMode() logic might be needed if fastmode is not enabled by default on ARM. Also, that logic assumes nearest is the default mode.

-[Unknown]

[hrydgard]

I don't have any numbers, but I doubt it's a lot, especially on newer cores than Cortex-A8 that don't have this "fast-mode" distinction..

[sum2012]

@hrydgard change 0x04086060 to 0x04c86060 don‘t help.

[sum2012]

My graphic probrem is more serious
Lg nexus 5
http://postimg.org/image/ue7adfka7/

[unknownbrackets]

How is it supposed to look?

If you change 0x03000000 to 0x00000000, it works? That will force nearest but keep FZ and DN off.

-[Unknown]

[sum2012]

It supposed to look http://postimg.org/image/lk0k0le2n/
I am at work so that I cannot compile

[unknownbrackets]

Hmm... I guess depth testing related?

-[Unknown]

[sum2012]

Revert hrydgard/native@3139ac4 fix it

[xiushudongfang]

@unknownbrackets change 0x03000000 to 0x00000000 ,works well.

[sum2012]

@unknownbrackets change 0x03000000 to 0x00000000 ,works well in my case too

[sum2012]

@hrydgard ,revert hrydgard/native@3139ac4
or use @unknownbrackets fix ?

[hrydgard]

With that, we no longer try to get arm cpus into "fastmode" globally. Should help.

Later, might add some smarter tracking of what flags we've set, so we can still use fastmode for things like software T&L etc, but not sure there's much of a win, really.

[sum2012]

@hrydgard Thanks,fixed grahic ,test in v1.0.1-777-gf4e0034.
Remain "infinite ingame loading screen" issue relate to atrac problem

[krispra]

sorry for my bad english... i hope you can understand what i mean :)

i've tried almost every build old build & lates build & i found 1 the best for playing tenchu sqn/ kurenai on android

i use ppsspp gold v0.7.6 to play tenchu san & tenchu kurenai on my android tab... and it work great there are no character through the wall / infinite loading screen both in tenchu san/ kurenai... but NO sound

[sum2012]

@krispra no character through the wall has been fixed in v1.0.1-772
But infinite loading screen have not been fixed.

This is hack version that fix these https://drive.google.com/file/d/0B3OaSdeV0L8kY3VXaXYyNkRRdmM/view?usp=sharing

[sum2012]

The reamin infinite ingame loading screen issue is same as #6370
I think that can close this

[unknownbrackets]

Thanks.

-[Unknown]

[Trokinos]

A good update about that bug, after noticing an user reporting that the "eternal loading" bug is not present anymore for Tenchu Kurenai in recent PPSSPP builds
http://forums.ppsspp.org/showthread.php?tid=14244&pid=112549#pid112549

And as it's the same engine as Tenchu San, i gave a try with my current build from a few days ago to my Tenchu San portable.

And i was able to complete the 1st mission of the Rikimaru path without a problem, the mission has several location in which there's a loading, but PPSSPP does not get stuck anymore on any of them (previously game wasn't playable, as soon as a loading time was happening, PPSSPP was getting stuck).
I don't have time to play more mission (as they're rather long) , but it looks like that game breaking bug is now fixed ! Thanks.
Not sure which commit did it, the version i am playing is :
1.1.1-273-g419ea84
on win32

[aquilsamuel]

This is what I get when playing this game, even on version 1.1.1 on the PPSSPP for windows 8 32 bit. The music still plays but this loading screen just freezes the video and nothing can be done. I am a Tenchu fan and would love to play and have this game really bad. Can anyone solve this annoying problem??? Thanks

[daniel229]

Download the latest build from http://buildbot.orphis.net/ppsspp/index.php

[aquilsamuel]

@daniel229 Thank you for your response and help, the new version of PPSSPP runs the game well and I was able to finish the first stage without any loading bugs.

[kedamtac]

thankkkkkkkkkkkk